Regional Convergence in Russia: Geografically Weighted Regression Approach

Ostrovskaya E., Mamontov D.

The main purpose of the paper is testing of the hypothesis about the correlation between the growth rate and initial level of per capita income (β-convergence). Geographically weighted regression (GWR) was chosen as the analysis tool due to significant spatial heterogeneity of parameters. We compare weighing matrices of distances, interregional trade and migration flows. Optimal bandwidth is determined using Akaike (AIC) and cross-validation (CV) criteria.

The paper shows the reasonability of separating Russian regions into western and eastern and shows the presence of unconditional β-convergence and σ-convergence in western regions. Estimation of conditional β-convergence shows slower convergence in regions with high share of public sector. The use of matrices of trade and migration (instead of the distance matrix) allows to improve the results, however the rate of convergence is low.

Regional Convergence in Russia: Geografically Weighted Regression